Date: Fri, 30 Oct 2020 22:30:13 -0500
From: Justin Pryzby <pryzby@telsasoft.com>
To: Michael Paquier <michael@paquier.xyz>
Cc: Tomas Vondra <tomas.vondra@2ndquadrant.com>,
	=?iso-8859-1?Q?Fabr=EDzio?= de Royes Mello <fabriziomello@gmail.com>,
	Tom Lane <tgl@sss.pgh.pa.us>,
	"David G. Johnston" <david.g.johnston@gmail.com>,
	pgsql-hackers@lists.postgresql.org,
	Nikolay Samokhvalov <samokhvalov@gmail.com>
Subject: Re: Add important info about ANALYZE after create Functional Index
Message-ID: <20201031033013.GE3080@telsasoft.com>
References: 
 <CAKFQuwb9GFe4oM3k9bsn7GOk5_jfow-9SpWocAG4xAYVp9nLjg@mail.gmail.com>
 <20201028185538.fzux2lhim472h46o@development>
 <CAKFQuwYGj-P3un2rGbj8Yc8oUohf_kUkDCnO5ap57bO85e56dQ@mail.gmail.com>
 <20201028191201.wh5ldvhsiqy2vknu@development>
 <1443542.1603912732@sss.pgh.pa.us>
 <20201028193546.ayzxua26od3qnbzt@development>
 <CAFcNs+qgLOZWJ=iugTME-UqcLOmGLVS_ioPpvGS87mJ=j8oLZw@mail.gmail.com>
 <20201028230211.dxbjw6mffq37mudu@development>
 <20201029015952.GB2671@paquier.xyz>
 <20201030062252.GG1693@paquier.xyz>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20201030062252.GG1693@paquier.xyz>
User-Agent: Mutt/1.9.4 (2018-02-28)
Precedence: bulk

On Fri, Oct 30, 2020 at 03:22:52PM +0900, Michael Paquier wrote:
> On Thu, Oct 29, 2020 at 10:59:52AM +0900, Michael Paquier wrote:
> > REINDEX CONCURRENTLY is by design wanted to provide an experience
> > transparent to the user similar to what a plain REINDEX would do, at
> > least that's the idea behind it, so..  This qualifies as a bug to me,
> > in spirit.
> 
> And in spirit, it is possible to address this issue with the patch
> attached which copies the set of stats from the old to the new index.
> For a non-concurrent REINDEX, this does not happen because we keep the
> same base relation, while we replace completely the relation with a
> concurrent operation.  We have a RemoveStatistics() in heap.c, but I
> did not really see the point to invent a copy flavor for this
> particular case.  Perhaps others have an opinion on that?

+1

The implementation of REINDEX CONCURRENTLY is "CREATE INDEX CONCURRENTLY
followed by internal index swap".  But the command is called "reindex", and so
the user experience is that the statistics are inexplicably lost.

(I'm quoting from the commit message of the patch I wrote, which is same as
your patch).

-- 
Justin